MACHINE LEARNING PROJECT

AUTO-MOBILE PRICE PREDICTION

1. BUSINESS CASE: THE AIM IS TO PREDICT THE PRICE OF CAR USING ALL THE GIVEN FEATURES

2. IMPORTING THE PYTHON LIBRARIES

3. LOADING THE DATASET

4. DOMAIN ANALYSIS

1. symboling: This rating corresponds to the degree to which the auto is more risky than its price indicates.

   • Cars are initially assigned a risk factor symbol associated with its price. Then, if it is more risky (or less), this symbol is adjusted by moving it up (or down) the scale.
   • Actuarians call this process "symboling". A value of +3 indicates that the auto is risky, -3 that it is probably pretty safe. (-3, -2, -1, 0, 1, 2, 3.)
     

2. normalized-losses: This factor is the relative average loss payment per insured vehicle year. This value is normalized for all autos within a particular size classification (two-door small, station wagons,sports/speciality, etc...), and represents the average loss per car per year. (continuous from 65 to 256)
   

3. make: This represents the maker/make of auto. (alfa-romero, audi, bmw, chevrolet, dodge, honda, isuzu, jaguar, mazda, mercedes-benz, mercury, mitsubishi, nissan, peugot, plymouth, porsche, renault, saab, subaru, toyota, volkswagen, volvo)
   

4. fuel-type: Type of fuel used in engine. (diesel, gas)
   

5. aspiration: If the auto has turbo or natually aspirated engine. (std, turbo)
   

6. num-of-doors: Number of doors in auto. (four, two)
   

7. body-style: Type/style of auto body. (hardtop, wagon, sedan, hatchback, convertible)
   

8. drive-wheels: The wheel drive system which transmits force causing the auto to move. (4wd, fwd, rwd)
   

9. engine-location: Placement of engine in auto. (front, rear)
   

10. wheel-base: The distance between the wheel axles - centers of front and rear wheels. (continuous from 86.6 120.9)
    

11. length: Length of auto. (continuous from 141.1 to 208.1)
    

12. width: Width of auto. (continuous from 60.3 to 72.3)
    

13. height: Height of auto. (continuous from 47.8 to 59.8)
    

14. curb-weight: Weight of the car with standard components. (continuous from 1488 to 4066)
    

15. engine-type: dohc(double overhead camp), dohcv, l, ohc, ohcf, ohcv(overhead camp valve), rotor.
    

16. num-of-cylinders: Number of cylinders used in auto. (eight, five, four, six, three, twelve, two)
    

17. engine-size: Size of auto engine. (continuous from 61 to 326)
    

18. fuel-system: The system that helps transfer fuel to the engine. (1bbl, 2bbl, 4bbl, idi, mfi, mpfi, spdi, spfi)
    

19. bore: Hollow part inside engine/inner diameter of the cylinder. (continuous from 2.54 to 3.94)
    
20. stroke: The full travel of piston along the cylinder. (continuous from 2.07 to 4.17)
    
21. compression-ratio: The ratio of maximum to minimum volume in the cylinder of an internal combustion engine. (continuous from 7 to 23
    
22. horsepower: Power that an engine produces. (continuous from 48 to 288)
    
23. peak-rpm: RPM that the engine produces at highest horsepower. (continuous from 4150 to 6600)
    
24. city-mpg: Lowest mpg rating for an auto. (continuous from 13 to 49)
    
25. highway-mpg: Highest/average mpg rating of an auto while driving on an open stretch road. (continuous from 16 to 54)
    
26. price: Cost of auto. (continuous from 5118 to 45400)

BASIC CHECKS

5.EXPLORATORY DATA ANALYSIS(EDA)

5.1 UNIVARIATE ANALYSIS:

DATA INSIGHTS:

• More than 30% of cars have a risk value of 0 which means that its averagly safe to buy.
  
• 15% of the cars are of the make Toyota.
  
• 90% of cars have gas engine and only 10% have diesel fuel engine.
  
• 82% of cars have naturally aspirated engine and 12% have turbo engines.
  
• 56% of cars have 4 doors, 43% have 2 doors.
  
• More than 45% of cars are Sedan and around 34% are hatchbacks.
  
• Almost 60% of cars have forward wheel drive.
  
• Just 1% of cars have engine located in the rear end of car.
  
• For most of the cars, wheel base is around 95 to 100.
  
• Around 25% of cars are of length 170units. This feature shows almost normal distribution.
  
• Around 25% of cars are of width range 66 to 68units.
  
• Most of the cars are of height range 55 to 56units.
  
• 25% of cars have curb weight of 2500units.
  
• More than 70% of cars have ohc (OverHead Camp) kind of engine.
  
• Almost 80% of cars have 4 cylinders.
  
• 45% of car's engine size is 100.
  
• Around 45% of cars have mpfi kind of fuel system.
  
• Most of the car's bore is more than 3.8units.
  
• Most of the car's stroke length is more than 3.5units.
  
• 65% of cars have compression ratio of 10
  
• More than 18% of cars have a horsepower of 68hp.
  
• Most of the cars have peak rpm of 5500 or 4800.
  
• Around 30% of cars give a milege of 25 in city.
  
• More than 20% of cars give a milege in the range of 25 to 35 in highway.
  
• 40% of cars have a price range of 5 to 10k.

5.2 BIVARIATE ANALYSIS:

DATA INSIGHTS:

• Price of Jaguar cars is the highest.
  
• The price of diesel engine cars is higher than gas engine cars.
  
• Turbo engined cars are pricey than the naturally aspirated cars.
  
• Cars with 4 doors are costlier than 2 door cars.
  
• Convertible and Hardtop type of cars have higher price than other types.
  
• Reverse wheel drive cars have higher price than forward or 4 wheel drive cars.
  
• Rear engine cars are costlier than front engine cars.
  
• OverHead Camp Valve (ohcv) engine have higher price.
  
• 8 cylinder engine cars have highest price.
  
• mpfi fuel system car has the highest price.

DATA INSIGHTS:

• At an average normalized loss of 142, the price of cars is 35k.
  

• Cars with bore measuring 3.8units has the highest price.
  

• Cars with stroke length of 2.76units has the highest price.
  

• A horsepower of 184 is paid the highest price.
  

• Car with peak RPM of 5900 is the costliest.
  

• A car with symboling value of -1 has higher price than that with value of 3.
  

• Car with wheel base of 112 is paid highest.
  

• Car with length of 199.2, width of 72 and height of 55.4 is highly paid.
  

• Car with curb weight of 3715 is highly paid.
  

• Engine size of 304 has highest price.
  

• Compression ratio of 11.5 is paid more.
  

• A city mpg of 14 is sold at highest price.
  

• A highway mpg of 16 is sold at highest price.

  5.3 MULTIVARIATE ANALYSIS:

  

  

  • Data Insight:Marcedes-benz with gas fuel type has highest price
  • Data Insight: With increase in weight of vehicle, city mpg decreases.

  6.FEATURE ENGINERING/ DATA PREPROCESSING

  6.1 Checking for missing or null values

  6.2 Converting categorical data to numerical data

  6.3 Using Label Encoder

  6.4 HANDLING OUTLIERS

  

  7.FEATURE SELECTION

  7.1 Checking for correlation

• USE OF HEAT MAP:

8.MODEL BUILDING & EVALUATION

LINEAR REGRESSION:

• Training R2 accuracy using Linear Regression is: 86.53144036257537
• Testing R2 accuracy using Linear Regression is: 76.72938159080243
• Testing Adjusted R2 score is: 48.28751464622761
• MSE score is: 28470850.70003938
• RMSE score is: 5335.808345512363
• MAE score is: 3406.7596001032775

K-Neighbors Regressor:

• Training R2 accuracy using KNN Regression is: 82.50133136918122
• Testing R2 accuracy using KNN Regression is: 64.93533173518867
• Testing Adjusted R2 score is: 22.078514967085916
• MSE score is: 42900490.114146344
• RMSE score is: 6549.846571801995
• MAE score is: 3946.4146341463415

SVM:

• Training R2 accuracy using SVM Regression is: -9.315945709087071
• Testing R2 accuracy using SVM Regression is: -21.91990376994277
• Testing Adjusted R2 score is: -170.9331194887617
• MSE score is: 149165068.0080445
• RMSE score is: 12213.315193183402
• MAE score is: 7943.122270031737

DECISION TREE REGRESSION:

• Training R2 accuracy using Decision tree Regression is: 99.89831862870213
• Testing R2 accuracy using Decision tree Regression is: 89.4528958266559
• Testing Adjusted R2 score is: 76.56199072590202
• MSE score is: 12904041.609756097
• RMSE score is: 3592.219593754827
• MAE score is: 1950.3902439024391

GRADIENT BOOSTING:

• The training R2 accuracy using Gradient Boosting is: 99.22053302005716
• The testing R2 accuracy using Gradient Boosting is: 94.9681530034094
• Testing Adjusted R2 score is: 88.81811778535422
• MSE score is: 6156302.426786809
• RMSE score is: 2481.1897200308586
• MAE score is: 1656.7798100138448

XGBOOST REGRESSOR:

• The training R2 accuracy using XG Boost is: 99.89824021512386
• The testing R2 accuracy using XB Boost is: 92.59952491032328
• Testing Adjusted R2 score is: 83.5544998007184
• MSE score is: 9054242.465007683
• RMSE score is: 3009.026830223965
• MAE score is: 1908.5585580221036

RANDOM FOREST REGRESSOR

• The training R2 accuracy using Random Forest is: 96.98578103166696
• The testing R2 accuracy using Random Forest is: 86.21079174165337
• Testing Adjusted R2 score is: 69.35731498145194
• MSE score is: 16870651.33773875
• RMSE score is: 4107.389844869701
• MAE score is: 2489.5789674796747

AFTER HPYERPARAMETER TUNING

• Training R2 score after Hyperparameter tuning on Random forest is: 99.89831862870213
• Testing R2 score after Hyperparameter tuning on Random forest is: 94.451462613851
• Testing Adjusted R2 score is: 87.6699169196689
• MSE score is: 6788456.445239862
• RMSE score is: 2605.466646349529
• MAE score is: 1553.0564634146342

So far, Gradient Boosting and Random Forest algorithms have given the best scores:

GRADIENT BOOSTING

• The training R2 accuracy using Gradient Boosting is: 99.22053302005716
  

• The testing R2 accuracy using Gradient Boosting is: 94.9681530034094
  

• Testing Adjusted R2 score is: 88.81811778535422
  

• MSE score is: 6156302.426786809
  

• RMSE score is: 2481.1897200308586
  

• MAE score is: 1656.7798100138448
  

  RANDOM FOREST

• Training R2 score after Hyperparameter tuning on Random forest is: 99.89831862870213
  

• Testing R2 score after Hyperparameter tuning on Random forest is: 94.451462613851
  

• Testing Adjusted R2 score is: 87.6699169196689
  

• MSE score is: 6788456.445239862
  

• RMSE score is: 2605.466646349529
  

• MAE score is: 1553.0564634146342

  CROSS VALIDATION SCORES:

  [ 0.6154544 0.36697642 -0.85743858]

• Cross validation score of Linear Regression model is: 0.041664079961660404 [ 0.30571883 0.55311535 -0.69954503]

• Cross validation score of KNN model is: 0.05309638233246209 [-0.20935183 -0.2448066 -0.01494369]

• Cross validation score of SVR model is: -0.15636737650881907 [0.35543453 0.4146508 0.12383546]

• Cross validation score of Decision Tree model is: 0.2979735956409533 [0.56014578 0.65703396 0.39638363]

• Cross validation score of Gradient Boost model is: 0.5378544577413296 [0.62451125 0.61963064 0.65658258]

• Cross validation score of XG Boost model is: 0.6335748203051849 [0.76522458 0.62552531 0.63047936]

• Cross validation score of Random Forest model is: 0.6737430824341969

• Training R2 score after Hyperparameter tuning on Random forest is: 99.89831862870213

• Testing R2 score after Hyperparameter tuning on Random forest is: 94.451462613851

• Testing Adjusted R2 score is: 87.6699169196689

• MSE score is: 6788456.445239862

• RMSE score is: 2605.466646349529

• MAE score is: 1553.0564634146342

• Cross validation score of Random Forest model is: 0.6737430824341969

  

  RESULT: Since Random Forest model comparitively has higher train, test and Cross validation scores and lower MSE, RMSE, MAE score, we choose this model for this problem